Building and managing cohesive interaction for virtual assistants

Abstract

A method includes receiving data comprising a plurality of requests and a plurality of responses to the requests. The requests and the responses are associated with a virtual assistant programmed to address the plurality of requests. In the method, a machine learning (ML) classifier is used to partition the requests into a plurality of partitions corresponding to a plurality of request types. An interface for a user is generated to display a subset of the requests corresponding to at least one partition of the plurality of partitions and to display a response corresponding to the subset of the plurality of requests, wherein the response is based on one or more of the plurality of responses. The interface is configured to permit editing of the response by the user. The method also includes processing the response edited by the user, and transmitting the edited response to the virtual assistant.

What is claimed is: 
     
       1. An apparatus comprising:
 at least one processing platform comprising a plurality of processing devices; 
 said at least one processing platform being configured: 
 to receive data comprising a plurality of requests and a plurality of responses to the plurality of requests, wherein the plurality of requests and the plurality of responses are associated with a virtual assistant programmed to address the plurality of requests; 
 to use a machine learning (ML) classifier to partition the plurality of requests into a plurality of partitions corresponding to a plurality of request types; 
 to generate an interface for a user to display a subset of the plurality of requests corresponding to at least one partition of the plurality of partitions and to display a response corresponding to the subset of the plurality of requests; 
 wherein the response is based on one or more of the plurality of responses; 
 wherein the interface is configured to permit editing of the response by the user; 
 to process the response edited by the user; 
 to transmit the edited response to the virtual assistant and 
 to use the ML classifier to execute an evaluation metric to compute a cohesion score for the plurality of requests and for the plurality of responses for each of a plurality of topics; 
 wherein the cohesion score measures a degree of correspondence between given requests of the plurality of requests and given responses of the plurality of responses for a given topic. 
 
     
     
       2. The apparatus of  claim 1  wherein said at least one processing platform is further configured to compute a frequency of usage of the plurality of requests. 
     
     
       3. The apparatus of  claim 2  wherein said at least one processing platform is further configured to rank the plurality of topics based on the cohesion score and the frequency of usage. 
     
     
       4. The apparatus of  claim 3  wherein in ranking the plurality of topics, said at least one processing platform is configured to rank topics with a first cohesion score and a first frequency of usage higher than topics with a second cohesion score and a second frequency of usage, wherein the first cohesion score is lower than the second cohesion score, and the first frequency of usage is higher than the second frequency of usage. 
     
     
       5. The apparatus of  claim 1  wherein the at least one partition further corresponds to a topic of the plurality of topics. 
     
     
       6. The apparatus of  claim 5  wherein said at least one processing platform is further configured to generate a new topic in addition to the plurality of topics by clustering a plurality of edited responses. 
     
     
       7. The apparatus of  claim 1  wherein, in partitioning the plurality of requests into the plurality of partitions, said at least one processing platform is configured to use the ML classifier to identify which of the plurality of request types correspond to a topic of the plurality of topics. 
     
     
       8. The apparatus of  claim 7  wherein said at least one processing platform is further configured to use an additional ML classifier to assign each of the plurality of requests to respective topics of the plurality of topics. 
     
     
       9. The apparatus of  claim 8  wherein said at least one processing platform is further configured to train the additional ML classifier with training data labeled by the respective topics of the plurality of topics. 
     
     
       10. The apparatus of  claim 1  wherein said at least one processing platform is further configured to train the ML classifier with training data labeled by respective ones of a plurality of request types. 
     
     
       11. The apparatus of  claim 1  wherein the virtual assistant comprises a chatbot. 
     
     
       12. An apparatus comprising:
 at least one processing platform comprising a plurality of processing devices; 
 said at least one processing platform being configured: 
 to receive data comprising a plurality of requests and a plurality of responses to the plurality of requests, wherein the plurality of requests and the plurality of responses are associated with a virtual assistant programmed to address the plurality of requests; 
 to use a machine learning (ML) classifier to partition the plurality of requests into a plurality of partitions corresponding to a plurality of request types; 
 to generate an interface for a user to display a subset of the plurality of requests corresponding to at least one partition of the plurality of partitions and to display a response corresponding to the subset of the plurality of requests; 
 wherein the response is based on one or more of the plurality of responses; 
 wherein the interface is configured to permit editing of the response by the user; 
 to process the response edited by the user; 
 to transmit the edited response to the virtual assistant 
 to determine a threshold number of the plurality of requests having a best fit to a request type of the at least one partition; and 
 to assign the threshold number of the plurality of requests to the subset. 
 
     
     
       13. The apparatus of  claim 1  wherein said at least one processing platform is further configured to automatically modify one or more of the plurality of responses based on a given one of the plurality of request types. 
     
     
       14. A method comprising:
 receiving data comprising a plurality of requests and a plurality of responses to the plurality of requests, wherein the plurality of requests and the plurality of responses are associated with a virtual assistant programmed to address the plurality of requests; 
 using a machine learning (ML) classifier to partition the plurality of requests into a plurality of partitions corresponding to a plurality of request types; 
 generating an interface for a user to display a subset of the plurality of requests corresponding to at least one partition of the plurality of partitions and to display a response corresponding to the subset of the plurality of requests; 
 wherein the response is based on one or more of the plurality of responses; 
 wherein the interface is configured to permit editing of the response by the user; 
 processing the response edited by the user; 
 transmitting the edited response to the virtual assistant; and 
 using the ML classifier to execute an evaluation metric to compute a cohesion score for the plurality of requests and for the plurality of responses for each of a plurality of topics; 
 wherein the cohesion score measures a degree of correspondence between given requests of the plurality of requests and given responses of the plurality of responses for a given topic; and 
 wherein the method is performed by at least one processing platform comprising at least one processing device comprising a processor coupled to a memory. 
 
     
     
       15. The method of  claim 14  further comprising computing a frequency of usage of the plurality of requests. 
     
     
       16. The method of  claim 15  further comprising ranking the plurality of topics based on the cohesion score and the frequency of usage. 
     
     
       17. The method of  claim 14  wherein partitioning the plurality of requests into the plurality of partitions comprises using the ML classifier to identify which of the plurality of request types correspond to a topic of the plurality of topics. 
     
     
       18. The method of  claim 17  further comprising using an additional ML classifier to assign each of the plurality of requests to respective topics of the plurality of topics. 
     
     
       19. The method of  claim 18  further comprising training the additional ML classifier with training data labeled by the respective topics of the plurality of topics. 
     
     
       20. The method of  claim 14  further comprising training the ML classifier with training data labeled by respective ones of a plurality of request types. 
     
     
       21. The method of  claim 14  further comprising automatically modifying one or more of the plurality of responses based on a given one of the plurality of request types. 
     
     
       22. A computer program product comprising a non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing platform causes said at least one processing platform:
 to receive data comprising a plurality of requests and a plurality of responses to the plurality of requests, wherein the plurality of requests and the plurality of responses are associated with a virtual assistant programmed to address the plurality of requests; 
 to use a machine learning (ML) classifier to partition the plurality of requests into a plurality of partitions corresponding to a plurality of request types; 
 to generate an interface for a user to display a subset of the plurality of requests corresponding to at least one partition of the plurality of partitions and to display a response corresponding to the subset of the plurality of requests; 
 wherein the response is based on one or more of the plurality of responses; 
 wherein the interface is configured to permit editing of the response by the user; 
 to process the response edited by the user; 
 to transmit the edited response to the virtual assistant; and 
 to use the ML classifier to execute an evaluation metric to compute a cohesion score for the plurality of requests and for the plurality of responses for each of a plurality of topics; 
 wherein the cohesion score measures a degree of correspondence between given requests of the plurality of requests and given responses of the plurality of responses for a given topic. 
 
     
     
       23. The computer program product of  claim 22  wherein the program code further causes said at least one processing platform to compute a frequency of usage of the plurality of requests. 
     
     
       24. The computer program product of  claim 23  wherein the program code further causes said at least one processing platform to rank the plurality of topics based on the cohesion score and the frequency of usage. 
     
     
       25. The computer program product of  claim 22  wherein, in partitioning the plurality of requests into the plurality of partitions, the program code further causes said at least one processing platform to use the ML classifier to identify which of the plurality of request types correspond to a topic of the plurality of topics. 
     
     
       26. The computer program product of  claim 25  wherein the program code further causes said at least one processing platform to use an additional ML classifier to assign each of the plurality of requests to respective topics of the plurality of topics. 
     
     
       27. The computer program product of  claim 26  wherein the program code further causes said at least one processing platform to train the additional ML classifier with training data labeled by the respective topics of the plurality of topics. 
     
     
       28. The computer program product of  claim 22  wherein the program code further causes said at least one processing platform to train the ML classifier with training data labeled by respective ones of a plurality of request types. 
     
     
       29. The computer program product of  claim 22  wherein the program code further causes said at least one processing platform to automatically modify one or more of the plurality of responses based on a given one of the plurality of request types.